That's not quite the same, though notifications and spam through the app would count.

The corpo LLMs do have a huge sycophancy bias, which would contribute as an addictive effect, I guess (and get people down strange psycological rabbit holes like you see in the news). I'd suggest switching between different models, especially the 'niche' open-weights finetunes on OpenRouter and such, that aren't so sycophantic. And also aren't OpenAI, because Altman is literally a blood sucking con artist.